Benchmarked 5 regression models (OLS, Ridge, Lasso, Random Forest, Gradient Boosting) on 1,200 used car listings, achieving R² = 0.952 and 5-fold CV R² = 0.948 using Python (scikit-learn, pandas) Reduced prediction error by 21.4% over baseline through 5 engineered features (power-to-displacement ratio, age-squared depreciation), delivering RMSE ₹10.65L and MAPE 24.95% on test Conducted EDA, multicollinearity diagnostics and residual analysis with visualization (Q-Q plots etc) using matplotlib and seaborn, with engine power and brand segment as top pricing drivers